Proceedings of the Annual Meeting of the Cognitive Science Society

This paper considers the issue why more pieces of supporting evidence may be less persuasive than fewer, perhaps to the point of dissuading one's interlocutor. Challenging the view that such more-is-less effects result from failures to make correct inferences based on the evidence, this paper argues that they may be a rational response based on one's beliefs about a source's reliability. To do this, we present a Bayesian network framework for inference with potentially unreliable sources. While existing Bayesian models forbid more-is-less effects, our more general framework naturally accounts for them.